Apparent Temperature Modifies the Effects of Air Pollution on Cardiovascular Disease Mortality in Cape Town, South Africa

Cardiovascular disease (CVD) is the top cause of mortality and a main contributor to disability globally [1]. During the past 30 years, CVD mortality declined in high-income countries; however, there is evidence that the reduction is stalling [2]. In contrast, CVD mortality has increased in low- and middle-income countries (LMIC) over the same period [1]. In South Africa, almost one in five natural deaths were due to CVD from 2016 to 2018 [3]. However, knowledge on the incidence and prevalence of CVD morbidity in South Africa is poor [4], and thus mortality is the tip of the proverbial iceberg. In addition, people in Africa have a poor understanding of CVD, its risk factors and clinical symptoms, which contributes to CVD mortality [5,6,7].

Evidence is increasing on the short- and long-term CVD effects of outdoor air pollution, one of the many risk factors of CVD [1,8,9,10,11]. There are thousands of chemicals that people can be exposed to via inhalation, however, the vast majority of epidemiological studies investigated particulate matter with an aerodynamic diameter less than or equal to 2.5 µm (PM_2.5), PM₁₀, ground-level ozone (O₃), nitrogen oxide (NO₂), sulphur dioxide (SO₂), carbon monoxide (CO) and lead [11].

Government air pollution monitoring [12,13,14] as well as exposure assessment and epidemiological studies are lacking in Africa [14,15,16], and the true burden of air pollution on human health in Africa is up for debate. A recent meta-analysis summarised the evidence from epidemiological studies that investigated the effects of air pollution on all-cause and cause-specific mortality, such as CVD, during 1992 to 2019 [10]. Only 44 studies were included in the meta-analysis that investigated the association between CVD mortality and PM₁₀. Studies are lacking for the other common air pollutants and also in LMIC [10]. Few studies investigated susceptibility by sex and age or two-pollutant models [10]. The meta-analysis included only one study from Africa. The latest WHO guideline report also highlighted that more research is needed on the human health effects of NO₂ and SO₂ [11].

Human activities that lead to air pollution emissions also contribute to climate change. Climate change will alter weather conditions such as temperature, humidity, wind speed, atmospheric direction and mixing height [17]. These weather conditions in turn may modify the human health effects of air pollution by altering levels, transboundary movement and chemical transformation and composition [18].

Global rising ambient temperature, one of the key climatic change indicators, is a cause for concern [19,20]. Ambient temperature has direct and indirect effects on CVD [21,22]. Epidemiological studies in Africa and other LMICs on the human health effects of heat and cold are lacking [23,24,25]. A review highlighted that LMIC populations are more susceptible to health effects of heat and cold [25]. Possible reasons for this include inadequate healthcare services, infrastructure and technology and poverty [25]. It is projected that by 2080, temperature increases greater than 4 °C will be observed across South Africa [26].

The majority of epidemiological studies that investigated air pollution as a risk factor for human health adjusted for ambient temperature as a confounder in regression models. However, fewer studied how the effects of air pollution on human health are modified by temperature [27,28]. Studies report inconsistent results, and there is a paucity of studies in Africa and other LMICs [29,30].

Mwase et al. compared PM₁₀, NO₂ and SO₂ levels from various African and global cities [43]. PM₁₀, NO₂ and SO₂ levels in Cape Town during 2006–2015 were similar or even lower than those in cities from developed countries, and markedly lower than those observed in two large South African cities (Durban and Johannesburg), industrial areas in the country or other LMICs in North Africa, Asia and South America [43].

Ischemic heart disease was the most common type of CVD death in Cape Town. Ischemic heart diseases are characterised by narrowed or blocked heart arteries that lead to suboptimal supply of oxygenated blood [44]. The synergistic effects of temperature and air pollution on CVD have a biological plausibility, such as increased clotting risk due to systemic inflammation after air pollution exposure, along with a rise in cholesterol levels and blood viscosity at high ambient temperature [28].

On days with low Tapp levels, significantly higher median NO₂ and SO₂ levels were observed compared to days with moderate or high Tapp levels, although Cape Town is situated in a winter rainfall region [31]. In contrast, the median PM₁₀ level was significantly higher on days with high Tapp levels. However, it is probable that during the 10-year study period, drier or wetter months also had days with low and high Tapp levels, respectively. Another explanation for the observed results is that of atmospheric chemistry, as discussed for NO₂ below.

Possible explanations for the observed PM₁₀ and NO₂ effects on moderate and high Tapp days may be due to human behaviour, physiology, toxicity of the air pollutants or meteorological factors such as rain or wind. On days with moderate and high Tapp levels people may open windows more, which may result in more outdoor air pollution infiltrating indoors. People may also participate in outdoor activities on such days and this may increase their exposure levels. The source and composition of air pollution may be influenced by ambient temperature. Studies observed that at higher temperature levels higher levels of toxic forms of PM tend to be detected [27,28]. Temperature may thus be an indicator of PM₁₀ composition.

A meta-analysis reported that CVD mortality significantly increased by 0.5%, 0.5% and 1.6% at low, moderate and high temperature, respectively, per 10 μg·m⁻³ increase in PM₁₀. The meta-analysis included 29 studies, mainly from China, Europe, USA, Australia and India [27]. No results were reported for two-pollutant models in the meta-analysis and no study from Africa was available. The risk for CVD mortality was higher in our study due to exposure to PM₁₀ on days with low and moderate Tapp levels, namely 1.9% and 0.8%, respectively, but lower on days with high Tapp levels (1.3%).

PM₁₀ effects were attenuated in the two-pollutant models, although the pollutant had weak to moderate correlations with NO₂ and SO₂. The attenuation may not only be due to confounding but may be an indicator of the source-related constituent of PM₁₀, as was proposed by Qian et al. [45].

Evidence is lacking as to what extent temperature modifies the effects of NO₂ or SO₂ on CVD mortality [27,28]. Our observation of a 3.1% and 4.7% increase in CVD mortality per 10 μg·m⁻³ increase in NO₂ at low and moderate Tapp levels are higher than results from three Chinese and Taiwanese studies that were included in the meta-analysis [27], but lower at high Tapp (0.6%). The meta-analysis reported significant increases of 1.5%, 1.5% and 0.9% at low, moderate and high temperature, respectively, per 10 μg·m⁻³ increase in NO₂ [27].

We did not observe any significant risks for SO₂ at any Tapp level. Three studies from China found increased CVD mortality of 1.4% and 1.3% at low and moderate temperature, respectively, per 10 μg·m⁻³ increase in SO₂, and an insignificant increase of 0.4% at high temperature [27].

NO₂ can be converted to water-soluble nitrate in atmospheric chemical reactions, which can partition to the particulate phase and thereby generate PM [46]. This may be a possible explanation for the higher risks due to NO₂ exposure compared to PM₁₀ at low and moderate Tapp. A meta-analysis also reported similar results [27]. A meta-analysis reported that natural-cause mortality increased by 17% per 1 μg·m⁻³ increase in nitrate [47], which is much higher compared to a 0.4% increase per 10 μg·m⁻³ increase in PM₁₀ reported in another meta-analysis [10].

In general, the elderly and females were more vulnerable to air pollution exposure, especially at high and moderate Tapp levels. Reviews highlighted that there is a research need to investigate subgroups of the population [27,28,48]. The elderly tends to be frailer and have more co-morbidities and less physiological resilience [21,48,49]. The inhaled air pollution dose may be larger in females than males, as they have smaller lung tissue and trachea than males [50]. Females also sweat less, have a higher working metabolic rate and may have thicker subcutaneous fat that may weaken thermoregulation [51].

A meta-analysis reported no significant difference in the pooled effects of short lags (lag0, lag1 or lag0–1) of PM₁₀ on CVD mortality and those of longer lags of up to a week [10]. This means that the deaths of very frail individuals were not merely brought forward by a few days, so no harvesting effects were observed. Harvesting effects are observed if the risks decrease at longer lags of air pollutants. This indicates that the frail individuals are removed from the population at short lags, leaving fewer frail individuals to be at risk at longer lags [52]. The meta-analysis reported pooled effects for all age groups and sexes combined. Few studies investigated susceptibility by age or sex, as mentioned before. In the current study, it appears that harvesting effects were not uniform for the different air pollutants nor the different subgroups. It was expected that harvesting effects will be clearly observed for females and the elderly, as they were identified to be more at risk from air pollution exposure at lag0–1. Possible reasons for the difference in harvesting effects among males and females may be due to physiology, as mentioned above, or exposure to indoor air pollution such as environmental tobacco smoke or polluting household fuel use for cooking and heating. Clearly, studies are needed in South Africa that investigate air pollution harvesting effects on mortality and hospital admissions. Even though harvesting may be observed in our study, from a public health risk communication perspective, it is beneficial to decrease air pollution levels in Cape Town. Chronic air pollution exposure leads to disease incidence [8,11], which in turn makes individuals in the population frail.

Air pollution and meteorological levels were measured at a few sites, and it was assumed that these levels are the same across the entire city. This may have led to measurement error. All ecological epidemiological studies have this limitation. However, it was shown that this exposure misclassification is non-differential and bias the effect estimates towards the null [53].

Another limitation is that PM_2.5 was not investigated in this study. The pollutant was not monitored in the City of Cape Town during 2006–2015. PM_2.5 can penetrate deeper into the respiratory tract than PM₁₀, penetrate the lung barrier and enter the blood system, thus making it more hazardous to human health than PM₁₀ [9,11].

The findings of this study can be applied to develop an early warning system for the city, e.g., use of machine learning methods [54]. For example, days can be grouped based on median daily air pollution levels at low, moderate and high Tapp levels. This can be a predictor variable for CVD mortality. It is important to note that the results cannot be extrapolated to other health outcomes such as hospital admissions, other cause-specific mortalities or to other cities in South Africa.